DOI: 10.1126/science.1176210 , 1110 (2009); 325Science

et al.Leo Gross,Atomic Force MicroscopyThe Chemical Structure of a Molecule Resolved by

This copy is for your personal, non-commercial use only.

. clicking herecolleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

. herefollowing the guidelines can be obtained byPermission to republish or repurpose articles or portions of articles

(this information is current as of September 13, 2010 ):The following resources related to this article are available online at www.sciencemag.org

http://www.sciencemag.org/cgi/content/full/325/5944/1110version of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/cgi/content/full/325/5944/1110/DC1 can be found at: Supporting Online Material

http://www.sciencemag.org/cgi/content/full/325/5944/1110#otherarticles, 4 of which can be accessed for free: cites 25 articlesThis article

18 article(s) on the ISI Web of Science. cited byThis article has been

http://www.sciencemag.org/cgi/collection/chemistryChemistry

: subject collectionsThis article appears in the following

registered trademark of AAAS. is aScience2009 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

only appearswhen looking at the Imix signal, its highQ is only an effective quality factor, but it is usefulfor applications that, for instance, rely on precisemeasurements of the resonance frequency (21).

These results show that the coupling betweenthe mechanics and the electron transport can bevery strong in SWNTs. In comparison, exper-iments on microfabricated semiconductor reso-nators, which are coupled to metal single-electrontransistors, have not shown any oscillations of f0and Q (22). This difference probably arises fromthe much greater mass of these resonators ascompared with those of SWNTs, so that they aremuch less sensitive to the motion of individualtunneling electrons. A way to quantify thecoupling is by looking at the damping rate causedby the interaction between the mechanics and theelectron transport, ge−ph ¼ 2pf0=Q. Dampingrate can be useful for the evaluation of quantumelectromechanics phenomena in analogy to lasercooling of, for example, trapped ions (23). Wehave ge−ph as high as 3 × 106 Hz for nanotubes,which compares with ge−ph ~ 0.7 Hz for mac-roscopic atomic force microscopy cantilevers(24). As for single-electron transistors that aresuperconducting, ge−ph is expected to be enhanced(15). In this case, however, ge−ph measured onmicrofabricated resonators is below5×104Hz (12).

The strong coupling in nanotubes holds prom-ise for various applications.We have shown that itallows for a widely tunable nonlinearity of the res-onator dynamic.Nonlinearity of themotion is usefulfor sensitivity improvement of mass and force sens-ing (25), mechanical signal amplification (26, 27),noise squeezing (28), study of quantum behaviors ofmacroscopic systems (29), mechanical microwavecomputing (30), or energy harvesting via vibration-to-electricity conversion (31). We emphasize thatthe nonlinearity in nanotubes can be tuned by anelectronic means (by changingVDC

g ), which is con-venient for practical use. In comparison, the non-

linearity in previously studied nanoresonators comesfrom purely mechanical effects (which can be de-scribed by the Duffing equation). There, the non-linearity can bemodified by using strain, but this iscomplicated to do experimentally. The strong cou-pling in nanotubes could also be used for coolingmechanical oscillations to the ground state (32),which would open the possibility to study non-classical states at a mesoscopic scale. The abilityto electrically tune the coupling (withVDC

g ) wouldthen be very useful for quantum manipulation.

References and Notes1. V. Sazonova et al., Nature 431, 284 (2004).2. P. Poncharal, Z. L. Wang, D. Ugarte, W. A. Heer, Science

283, 1513 (1999).3. B. Reulet et al., Phys. Rev. Lett. 85, 2829 (2000).4. S. T. Purcell, P. Vincent, C. Journet, V. T. Binh, Phys. Rev.

Lett. 89, 276103 (2002).5. B. Witkamp, M. Poot, H. S. J. van der Zant, Nano Lett. 6,

2904 (2006).6. D. Garcia-Sanchez et al., Phys. Rev. Lett. 99, 085501 (2007).7. B. Lassagne, D. Garcia-Sanchez, A. Aguasca, A. Bachtold,

Nano Lett. 8, 3735 (2008).8. K. Jensen, K. Kim, A. Zettl, Nat. Nanotechnol. 3, 533 (2008).9. H.-Y. Chiu, P. Hung, H. W. C. Postma, M. Bockrath,

Nano Lett. 8, 4342 (2008).10. A. Eriksson et al., Nano Lett. 8, 1224 (2008).11. J.-C. Charlier, X. Blase, S. Roche, Rev. Mod. Phys. 79, 677

(2007).12. A. Naik et al., Nature 443, 193 (2006).13. Materials and methods are available as supporting

material on Science Online.14. M. Brink, thesis, Cornell University (2007).15. A. A. Clerk, S. Bennett, N. J. Phys. 7, 238 (2005).16. We took Cdot = 57 aF from Fig. 2A. Dissipation caused by

He gas used to thermalize the resonator was included bycalculating the total quality factor as ð∑1=QiÞ−1. Thecontribution of Q from He has been measured to be 1035by means of pumping the gas, which left the temperaturestable for a short moment.

17. We obtained Cg ′2=k = 6 × 10–22 F2/Nm, which hasto be compared with Cg ′≈10−13−10−12F/m obtained byusing commercial simulators and k ≈ 10−4−10−3N/mfrom (1). The value of G is 3 × 109 s−1, which issomewhat smaller than expected in the quantum regimeG ¼ 8kBTGmax=e2 ≈ 9 � 1010 s−1 (where Gmax is themaximum conductance of a Coulomb-blockade peak). In

addition, the shape of the measured oscillations of f0 andQ differs to some extent from what is predicted (Fig. 3).These differences may arise from the fact that the nanotubedot is not perfect (the Coulomb blockade peaks are notfully periodic in VDCg , and the peak heights are different).Another explanation is that the model that was used is toosimple to quantitatively capture the physics (33, 34).

18. The oscillation ofQ as a function ofVDCg has been observed in asecond device. The Coulomb-blockade oscillations of G are,however, less regular than the ones presented here, and so thenanotube probably consists of a series of quantum dots (35).

19. W. G. Conley, A. Raman, C. M. Krousgrill, S. Mohammadi,Nano Lett. 8, 1590 (2008).

20. L. D. Landau, E. M. Lifshitz, Mechanics (Pergamon Press,Oxford, 1960).

21. When VACg is low, the values of f0 and Q extracted fromImix( f ) are the same as the ones from dz( f ).

22. R. G. Knobel, A. N. Cleland, Nature 424, 291 (2003).23. C. Raab et al., Phys. Rev. Lett. 85, 538 (2000).24. M. T. Woodside, P. L. McEuen, Science 296, 1098 (2002).25. J. S. Aldridge, A. N. Cleland, Phys. Rev. Lett. 94, 156403

(2005).26. A. Erbe et al., Appl. Phys. Lett. 77, 3102 (2000).27. R. L. Badzey, P. Mohanty, Nature 437, 995 (2005).28. R. Almog, S. Zaitsev, O. Shtempluck, E. Buks, Phys. Rev.

Lett. 98, 078103 (2007).29. V. Peano, M. Thorwart, Phys. Rev. B 70, 235401 (2004).30. S. B. Shim, M. Imboden, P. Mohanty, Science 316, 95 (2007).31. F. Cottone, H. Vocca, L. Gammaitoni, Phys. Rev. Lett.

102, 080601 (2009).32. S. Zippilli, G. Morigi, A. Bachtold, Phys. Rev. Lett. 102,

096804 (2009).33. A. D. Armour, M. P. Blencowe, Y. Zhang, Phys. Rev. B 69,

125313 (2004).34. F. Pistolesi, S. Labarthe, Phys. Rev. B 76, 165317 (2007).35. B. Gao, D. C. Glattli, B. Placais, A. Bachtold, Phys. Rev. B

74, 085410 (2006).36. We thank G. Morigi and J. Moser for discussions. The

research has been supported by the European Union–fundedprojects FP6-IST-021285-2 and FP6-ICT-028158, and theSwedish Foundation for Strategic Research.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/1174290/DC1Materials and Methods

31 March 2009; accepted 19 June 2009Published online 23 July 2009;10.1126/science.1174290Include this information when citing this paper.

The Chemical Structure of a MoleculeResolved by Atomic Force MicroscopyLeo Gross,1* Fabian Mohn,1 Nikolaj Moll,1 Peter Liljeroth,1,2 Gerhard Meyer1

Resolving individual atoms has always been the ultimate goal of surface microscopy. The scanningtunneling microscope images atomic-scale features on surfaces, but resolving single atomswithin an adsorbed molecule remains a great challenge because the tunneling current is primarilysensitive to the local electron density of states close to the Fermi level. We demonstrate imagingof molecules with unprecedented atomic resolution by probing the short-range chemical forceswith use of noncontact atomic force microscopy. The key step is functionalizing the microscope’s tipapex with suitable, atomically well-defined terminations, such as CO molecules. Our experimentalfindings are corroborated by ab initio density functional theory calculations. Comparison withtheory shows that Pauli repulsion is the source of the atomic resolution, whereas van der Waals andelectrostatic forces only add a diffuse attractive background.

Noncontact atomic force microscopy(NC-AFM), usually operated in frequency-modulation mode (1), has become an

important tool in the characterization of nano-

structures on the atomic scale. Recently, impres-sive progress has been made, including atomicresolution with chemical identification (2) andmeasurement of the magnetic exchange force

with atomic resolution (3). Moreover, lateral (4)and vertical (5) manipulation of atoms and thedetermination of the vertical and lateral forcesduring such manipulations (6) have been dem-onstrated. Striking results have also been ob-tained in AFM investigations of single molecules.For example, atomic resolution was achieved onsingle-walled carbon nanotubes (SWNTs) (7, 8),and the force needed to switch a molecularconformation was measured (9).

However, the complete chemical structure ofan individual molecule has so far not been im-aged with atomic resolution. The reasons thatmake AFM investigations on single molecules sochallenging are the strong influence of the exact

1IBM Research, Zurich Research Laboratory, 8803 Rüschlikon,Switzerland. 2Debye Institute for Nanomaterials Science,Utrecht University, Post Office Box 80000, 3508 TA Utrecht,Netherlands.

*To whom correspondence should be addressed. E-mail:lgr@zurich.ibm.com

28 AUGUST 2009 VOL 325 SCIENCE www.sciencemag.org1110

REPORTS

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

atomic composition and the geometry of the tip,as well as the relatively low stability of the sys-tem, which can result in unintentional lateral orvertical manipulation of the molecule duringimaging. As will be shown below, both problemscan be solved by preparing a well-defined tip bydeliberately picking up different atoms and mole-cules with the tip apex. The exact knowledge ofthe tip termination also facilitates quantitative com-parison with first-principles calculations, whichis essential for understanding the nature of thetip-sample interaction.

To benchmark AFM resolution on molecules,we investigated pentacene (C22H14, Fig. 1A), awell-studied linear polycyclic hydrocarbon con-sisting of five fused benzene rings. State-of-the-art scanning tunneling microscopy (STM) studiesof pentacene on metal, such as Cu(111) (10), andthin-film insulators, such as NaCl on Cu(111)(11, 12), have been performed recently. On in-sulating films, STM was used to image the mo-lecular orbitals near the Fermi level, EF, whereason metals the molecular orbitals were broadenedand distorted because of coupling to the elec-tronic states of the substrate. STM is sensitive tothe density of states near EF, which extends overthe entire molecule. This prevents the direct im-aging of the atomic positions (or core electrons)in such planar aromatic molecules by STM. Inthis work, we present atomically resolved AFMmeasurements of pentacene both on a Cu(111)substrate and on a NaCl insulating film.

For atomic resolution with the AFM, it isnecessary to operate in the short-range regime offorces, where chemical interactions give substan-tial contributions. In this force regime, it is de-sirable to work with a cantilever of high stiffnesswith oscillation amplitudes on the order of 1Å, aspointed out byGiessibl (13). Our low-temperatureSTM/AFM has its basis in a qPlus sensor design(14) and is operated in an ultrahigh vacuum at atemperature of 5 K. The high stiffness of thetuning fork [spring constant k0 ≈ 1.8 × 103 N/m(15), resonance frequency f0 = 23,165 Hz, andquality factor Q ≈ 5 × 104] allows stable opera-tion at oscillation amplitudes down to 0.2 Å. Ametal tip (16) was mounted on the free prong ofthe tuning fork, and a separate tip wire (which isinsulated from the electrodes of the tuning fork)was attached tomeasure the tunneling current (17).The bias voltage V was applied to the sample.

Modification of the STM tip apex is knownto have a profound influence on the achievableimage resolution (10, 11, 18, 19). We exploredthe effects of controlled atomic-scale modifica-tion of the AFM tip and show that suitable tiptermination results in dramatically enhanced atomicscale contrast in NC-AFM imaging. We imagedpentacene molecules (Fig. 1A) in STM (Fig. 1B)andAFM (Fig. 1, C andD)modes on Cu(111) byusing a CO-terminated tip. For these measure-ments, a COmolecule was deliberately picked upwith the tip (16), which led to an increased resolu-tion in the AFMmode (see below). From previousinvestigations, it is known that the CO molecule is

adsorbed with the carbon atom toward the metaltip (18, 19).

The CO molecule slightly affects the STMimage, and several faint maxima and minimaare visible because of the interaction of the COwith the pentacene orbitals, similar to the effectof a pentacene-modified tip (10). The AFM im-ages (Fig. 1, C and D) were recorded in constant-height mode; that is, the tip was scanned withoutz feedback parallel to the surface while thefrequency shift Df was being recorded (16). Inthis and all of the following measurements, thetip height z is always given with respect to theSTM set point over the substrate. The use ofconstant-height operation was critical because itallowed stable imaging in the region where Df is anonmonotonic function of z. In the AFM images(Fig. 1, C andD), the five hexagonal carbon ringsof each pentacene molecule are clearly resolved.

We observed local maxima of Df(x, y) above theedges of the hexagons, near the carbon atompositions, and minima above the centers of thecarbon rings (hollow sites), in concordance to themeasurements on SWNTs (7). Even the carbon-hydrogen bonds are imaged, indicating the posi-tions of the hydrogen atoms within the pentacenemolecule. Additionally, each molecule is sur-rounded by a dark halo.

To demonstrate that imaging conditions arealso stable for the case of organic moleculeson insulators, we used a thin insulating layer[NaCl(2 ML)/Cu(111), that is, two atomic layersof NaCl on Cu(111)] as substrate (Fig. 2). Fur-thermore, to study the influence of the tiptermination, we performed measurements withdifferent atomic modifications of the tip apex. Inaddition to the Ag- (Fig. 2A) and CO-terminated(Fig. 2B) tips, we also recorded Df images with

B

C

1.3Å

0Å5Å

5Å

-2Hz

-7Hz

+1Hz

-5Hz20Å

D

5Å

A

Fig. 1. STM and AFM imaging of pentacene on Cu(111). (A) Ball-and-stick model of the pentacenemolecule. (B) Constant-current STM and (C and D) constant-height AFM images of pentacene acquiredwith a CO-modified tip. Imaging parameters are as follows: (B) set point I = 110 pA, V = 170 mV; (C) tipheight z = –0.1 Å [with respect to the STM set point above Cu(111)], oscillation amplitude A = 0.2 Å; and(D) z = 0.0 Å, A = 0.8 Å. The asymmetry in the molecular imaging in (D) (showing a “shadow” only on theleft side of the molecules) is probably caused by asymmetric adsorption geometry of the CO molecule atthe tip apex.

A B

C -1Hz

-9Hz

-0.8Hz

-3.5Hz

-0.8Hz

-3.5Hz

-1.5Hz

-8Hz

5Å

5Å

5Å

5Å

D

Fig. 2. Constant-height AFM images of pentacene on NaCl(2ML)/Cu(111) using different tip modifications(16). (A) Ag tip, z= –0.7 Å, A= 0.6 Å; (B) CO tip, z=+1.3 Å, A= 0.7 Å; (C) Cl tip, z= –1.0 Å, A= 0.7 Å; and(D) pentacene tip, z=+0.6 Å, A=0.5 Å. The z values are given with respect to a STM set point of I=2 pA, V=200 mV above the NaCl(2 ML)/Cu(111) substrate.

www.sciencemag.org SCIENCE VOL 325 28 AUGUST 2009 1111

REPORTS

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

tips modified with Cl (Fig. 2C) and pentacene(Fig. 2D) (16). For each tip, the tip height z wasminimized; that is, decreasing z further by a few0.1 Å resulted in unstable imaging conditions,

usually leading to the molecule being laterallydisplaced or picked up by the tip. Comparing thedifferent tips, we see that their atomic terminationis crucial for the contrast observed above the

molecule. The highest lateral resolution was ob-served with CO-modified tips, and the contrastabove pentacene is similar to that in the measure-ments on Cu(111). With metal-terminated tips

Fig. 3. Maps of measured frequency shift Df (A) and extracted vertical force Fz(B) at different tip heights z. Corresponding line profiles of Df (C) and Fz (D)along the long molecular axis. The data shown are part of a complete three-

dimensional force field that has been measured in a box of 25 Å by 12.5 Å by13 Å above a pentacene molecule (16). The z values are given with respect to aSTM set point of I = 2 pA, V = 200 mV above the NaCl(2 ML)/Cu(111) substrate.

Fig. 4. Calculated energy map (A) for a CO-pentacene distance of d =4.5 Å. Calculated line profiles of the energy (B), the vertical force (C),and Df (D) above the long molecular axis for different moleculardistances. Calculated (E) and measured (F) force-distance curves abovedifferent molecular sites: central hollow site (blue), C-C bond of central

ring on long molecular axis (red), C atom of central ring on shortmolecular axis (green), and H atom on short axis (orange). The inset in(E) shows a measured Df map with the different molecular sites indicated.In the experimental data, the force components that arise from the metaltip behind the CO molecule have been subtracted (16, 29).

28 AUGUST 2009 VOL 325 SCIENCE www.sciencemag.org1112

REPORTS

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

(Ag, Au, or Cu), the molecule would always bemanipulated (usually being picked up by the tipfrom NaCl) before the minimum of Df(z) wasreached, and no atomic resolution could beobserved in AFM measurements. The Cl termi-nation yielded a contrast similar to that of the COtip. However, in the AFM image acquired withthe Cl tip, the minima above the hollow sites areless pronounced and the carbon rings appearsmaller in diameter compared with the CO-terminated tip. The pentacene-modified tip gavea completely different contrast compared with allthe other tips investigated, which indicates thestrong influence of the tip modification.

In the following, we concentrate on the in-vestigation of pentacene on NaCl(2 ML)/Cu(111)with a CO-terminated tip. We describe how thecontrast depends on the tip sample distance, quan-tify the forces acting on the tip, and lastly com-pare themeasurements to density functional theory(DFT) calculations in order to separate the con-tributions of different forces and understand theorigin of the observed contrast. Frequency shiftand force versus distance relations were deter-mined by capturing a three-dimensional (x, y, z)field (16, 20, 21) of the frequency shift with80 by 40 by 3100 data points in a box of 25 Åby 12.5 Å by 13 Å above a pentacene molecule.We extracted the vertical forces Fz(x, y, z) withuse of the method of Sader and Jarvis (16, 22).Figure 3A shows the measured frequency shiftat different tip heights, and the extracted verticalforce is shown in Fig. 3B. Corresponding lineprofiles along the long molecular axis are givenin Fig. 3, C and D, for Df and Fz, respectively.

For distances greater than z = 4.2 Å (topimage in Fig. 3A), we recorded only relativelysmall long-range forces (Fz < 20 pN), and themolecule was imaged as a featureless depression(attractive interaction).With decreasing tip height,Df(z) decreased before reaching a minimum(of about –4 Hz) above the molecular center atz ≈ 1.8 Å. Near the minimum of Df(z), we startedobserving corrugation on the atomic scale (Fig.3C). When we decreased z further, Df increasedagain and finally became even positive oversome parts of the molecule at z = 1.2 Å. Thelateral contrast of Df generally increased withdecreasing tip height (Fig. 3C). At the height atwhich Df crossed zero (z ≈ 1.2 Å), we achievedthe highest contrast and lateral resolution withAFM. This is the regime of maximal attractiveforces (16). Decreasing the tip height furtherwould result in instabilities and lastly in pickingup the molecule by the tip.

In the force maps (Fig. 3B), we likewise ob-served how the contrast increased with decreas-ing z. For the smallest accessible tip height, thatis, z = 1.2 Å, we measured an attractive force of110 pN above the central carbon ring. Above thepositions of the carbon atoms, the absolute valuesof the forces were smaller (between 60 and 90 pN)(15). These values are comparable to themaximumshort-range forces acting on a silicon tip above thehollow sites and the carbon positions of a SWNT,

which were recently measured as 106 pN and75 pN, respectively (7) (also with a systematicerror of about 30%).

To further elucidate the origin of the observedatomic contrast, we carried out DFT calculations(16, 23, 24) with highly optimized plane-wavecode CPMD (25). We applied the PBE (Perdew-Burke-Ernzerhof) exchange-correlation function-al (26) and used ab initio norm-conservingpseudopotentials (27). Van der Waals (vdW)forces were added semiempirically to the disper-sion energy as a contribution proportional to R−6

(where R is the atomic distance) (28). We onlyincluded the pentacene molecule and the COmolecule in our calculations and neglected boththe substrate and the metallic part of the tip. Weassumed for the calculations that the COmolecule is perpendicular to the plane of thepentacene molecule, with the oxygen atompointing toward the pentacene (16).

The calculated interaction energy surface isgiven in Fig. 4A, and the calculated constant-height line profiles of the energy, the verticalforce, and Df along the central molecular axis areshown in Fig. 4, B, C, and D, respectively. Theintermolecular distance d denotes the distance ofthe CO carbon atom to the plane of the pentaceneatoms. Experiment and theory (compare Figs. 3Dand 4C) concordantly showed maximal attractiveforces on the order of 100 pN above the hollowsites of the pentacene molecule, whereas attract-ive forces above the C-C bonds were smaller. Thedifference in the forces measured above these twosites increased with decreasing tip height. In theshort-range regime (d < 5Å), differences betweencalculations and experiment can be observed. Inthe calculations, sharp peaks appear above theouter C-C bonds (x = 6.4 Å and x = –6.4 Å in Fig.4, C and D), which are less pronounced in themeasurements (Fig. 3, C and D).

For a comparison between theory and exper-iment, we have to estimate the contribution of themetallic tip behind the CO molecule that was notincluded in the calculations. For this purpose, wemeasured the force acting on a purely metallictip, then picked up a CO molecule and measuredthe force again under otherwise identical con-ditions. The difference of the forces for identicaltip positions yielded the estimated contributionfrom the COmolecule only.We observed that theCO contribution to the force predominates in therelevant regime, whereas the metallic part of thetip contributed only about 30% to the attractiveforces and gave no corrugation on the atomicscale (16). Calculated force-versus-distancecurves (Fig. 4E) above different molecular sitesaround the central carbon ring of the molecule arecompared with experimental data, with the con-tribution of the metallic tip subtracted (Fig. 4F).We observe good qualitative agreement betweentheory and experiment in terms of the maximalvalues of the force and the relative order inwhich the maximal attractive forces are reachedabove the different atomic sites (29). Quantita-tively, the agreement is excellent for d > 5 Å. In

the short-range regime (4.5 Å < d < 5 Å), thecalculations overestimate the difference in theforces above the C-C bond and the hollow site.These discrepancies in the short-range regionmight arise because of the simplifications as-sumed in the calculations, namely not taking thesubstrate or the tip behind the CO into accountand the semiempirical treatment of vdW forces.Another origin of the discrepancies could be theincreasing influence of noise in the experimentaldata for small z values.

The calculations take into account forces ofthree different physical origins, namely electro-static forces, vdW forces, and Pauli repulsiveforces. Comparing their contributions to the over-all force, we found that the electrostatic forces aresmall (~10%) compared with the vdW forces.These two contributions to the force show littlelateral corrugation on the atomic scale and yield adiffuse attractive potential above the entire mole-cule, giving rise to the observed dark halo sur-rounding the molecules in the Df maps. Theorigin of the atomic contrast is the Pauli repulsionforce, which becomes substantial when regionsof high electron density overlap. These regionsare concentrated to the atomic positions and tothe C-C (and to a lesser extent also to the C-H)bonds in the pentacene molecule and are re-vealed for sufficiently small tip-sample distances(d ≈ 5 Å).

We conclude that atomic resolution in NC-AFM imaging onmolecules can only be achievedby entering the regime of repulsive forces becausethe vdWand electrostatic forces only contribute adiffuse attractive background with no atomic-scale contrast. Modifying the tip with suitableatomic or molecular terminations is required toallow the AFM to be operated in this regimewhile maintaining stable imaging conditions. Thetip termination governs the AFM contrast, andexact knowledge of the tip is needed for a detailedinterpretation of the force measurements. In thecase of the CO-terminated tip, we observed aspectacular enhancement of the atomic-scale con-trast and were able to resolve the atomic positionsand bonds inside pentacene molecules, preciselyrevealing the atomic molecular structure. It mayalso be possible to extract details about inter-molecular bonds, for example, bond order andlength. Furthermore, we foresee probing thereactivity of different molecular sites with re-spect to the known molecule or atom at the tipapex. Such investigations could yield detailedinsight into chemical reactions and catalysis.Lastly, a combination of NC-AFM with elec-trostatic force microscopy could be used to in-vestigate single-electron transport and chargedistributions in metal-molecule systems on theatomic scale.

References and Notes1. T. Albrecht, P. Grütter, D. Horne, D. Rugar, J. Appl. Phys.

69, 668 (1991).2. Y. Sugimoto et al., Nature 446, 64 (2007).3. U. Kaiser, A. Schwarz, R. Wiesendanger, Nature 446, 522

(2007).

www.sciencemag.org SCIENCE VOL 325 28 AUGUST 2009 1113

REPORTS

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

4. Y. Sugimoto et al., Nat. Mater. 4, 156 (2005).5. Y. Sugimoto et al., Science 322, 413 (2008).6. M. Ternes, C. P. Lutz, C. F. Hirjibehedin, F. J. Giessibl,

A. J. Heinrich, Science 319, 1066 (2008).7. M. Ashino, A. Schwarz, T. Behnke, R. Wiesendanger,

Phys. Rev. Lett. 93, 136101 (2004).8. M. Ashino et al., Nat. Nanotechnol. 3, 337 (2008).9. Ch. Loppacher et al., Phys. Rev. Lett. 90, 066107 (2003).

10. J. Lagoute, K. Kanisawa, S. Fölsch, Phys. Rev. B 70,245415 (2004).

11. J. Repp, G. Meyer, S. M. Stojkovic, A. Gourdon,C. Joachim, Phys. Rev. Lett. 94, 026803 (2005).

12. J. Repp, G. Meyer, S. Paavilainen, F. E. Olsson,M. Persson, Science 312, 1196 (2006).

13. F. J. Giessibl, Rev. Mod. Phys. 75, 949 (2003).14. F. J. Giessibl, Appl. Phys. Lett. 76, 1470 (2000).15. Mainly because of the uncertainty in the spring constant

of the cantilever, a systematic error of 30% is estimatedfor the measured forces.

16. Materials and methods are available as supportingmaterial on Science Online.

17. M. Heyde, M. Sterrer, H.-P. Rust, H.-J. Freund, Appl. Phys.Lett. 87, 083104 (2005).

18. L. Bartels et al., Phys. Rev. Lett. 80, 2004 (1998).19. H. J. Lee, W. Ho, Science 286, 1719 (1999).20. H. Hölscher, S. M. Langkat, A. Schwarz, R. Wiesendanger,

Appl. Phys. Lett. 81, 4428 (2002).21. B. J. Albers et al., Nat. Nanotechnol. 4, 307 (2009).22. J. E. Sader, S. P. Jarvis, Appl. Phys. Lett. 84, 1801 (2004).23. P. Hohenberg, W. Kohn, Phys. Rev. 136, B864 (1964).24. W. Kohn, L. J. Sham, Phys. Rev. 140, A1133 (1965).25. CPMD, IBM Corporation and MPI für Festkörperforschung

Stuttgart, www.cpmd.org/.26. J. P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77,

3865 (1996).27. D. R. Hamann, Phys. Rev. B 40, 2980 (1989).28. S. Grimme, J. Comput. Chem. 27, 1787 (2006).29. The experimental and theoretical distance values z and d

could be related to each other, for example, by comparingthe position of the force minimum above a certain molecularsite. However, the experimental z position of the forceminimum varies significantly between measurements withdifferent CO tips (in Fig. 3 the force minimum was alreadyreached at z ≈ 1.2 A, whereas in Fig. 4F it is not reacheduntil z ≈ 0.4 A). We attribute this to the fact that thetunneling current, and hence the tip height corresponding

to the STM set point, strongly depends on the exactadsorption geometry of the CO molecule at the tip apex.Therefore, we abstain from giving an explicit relationbetween z and d.

30. We thank A. Curioni, J. Repp, and R. Allenspach for valuablediscussions and comments and M. Heyde and F. J. Giessiblfor fruitful discussions on instrumental issues. The researchleading to these results has received funding from the SwissNational Center of Competence in Research (NCCR)“Nanoscale Science” and from the European Community’sSeventh Framework Programme under grant agreement no.214954 (HERODOT). P.L. gratefully acknowledges fundingby the Nederlandse Organisatie voor WetenschappelijkOnderzoek (Chemical Sciences, Vidi-grant 700.56.423).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/325/5944/1110/DC1Materials and MethodsFigs. S1 to S5References

12 May 2009; accepted 2 July 200910.1126/science.1176210

Amplifying the Pacific Climate SystemResponse to a Small 11-Year SolarCycle ForcingGerald A. Meehl,1* Julie M. Arblaster,1,2 Katja Matthes,3,4 Fabrizio Sassi,5 Harry van Loon1,6

One of the mysteries regarding Earth’s climate system response to variations in solar output ishow the relatively small fluctuations of the 11-year solar cycle can produce the magnitudeof the observed climate signals in the tropical Pacific associated with such solar variability.Two mechanisms, the top-down stratospheric response of ozone to fluctuations of shortwavesolar forcing and the bottom-up coupled ocean-atmosphere surface response, are included inversions of three global climate models, with either mechanism acting alone or both actingtogether. We show that the two mechanisms act together to enhance the climatologicaloff-equatorial tropical precipitation maxima in the Pacific, lower the eastern equatorialPacific sea surface temperatures during peaks in the 11-year solar cycle, and reducelow-latitude clouds to amplify the solar forcing at the surface.

It has long been noted that the 11-year cycleof solar forcing is associated with variousphenomena in Earth’s climate system, in both

the troposphere and stratosphere (1–9). Becausethe amplitude of the solar cycle (solar maxi-mum to solar minimum) is relatively small, about0.2 W m−2 globally averaged (10), and the ob-served global sea surface temperature (SST) re-sponse of about 0.1°C would require more than0.5Wm−2 (11), there has always been a questionregarding how this small solar signal could beamplified to produce a measurable response.

Postulatedmechanisms that could amplify therelatively small solar forcing signal to producesuch responses in the troposphere include changesin clouds in the troposphere caused by galacticcosmic rays, or associated global atmosphericelectric circuit variations, though neither has beenplausibly simulated in a climatemodel. However,there are two other plausible mechanisms, thougheach has not yet produced a modeled response ofthe magnitude seen in the observations. The firstinvolves a “top down” response of stratosphericozone to the ultraviolet (UV) part of the solarspectrum that varies by a few percent. Peaks insolar forcing cause the enhanced UV radiation,which stimulates additional stratospheric ozoneproduction andUVabsorption, thus warming thatlayer differentially with respect to latitude. Theanomalous temperature gradients provide a posi-tive feedback throughwavemotions to amplify theoriginal solar forcing. The changes in thestratosphere modify tropical tropospheric circu-lation and thus contribute to an enhancement andpoleward expansion of the tropical precipitation

maxima (5, 12–16). The first demonstration ofthe top-down mechanism in a modeling studyshowed a broadening of the Hadley cells inresponse to enhanced UV that increased as thesolar-induced ozone change was included (17).

A second “bottom up” mechanism that canmagnify the response to an initially small solarforcing involves air-sea coupling and interactionwith incoming solar radiation at the surface in therelatively cloud-free areas of the subtropics. Thus,peaks in solar forcing produce greater energy inputto the ocean surface in these areas, evaporatingmore moisture, and that moisture is carried by thetrade winds to the convergence zones where moreprecipitation occurs. This intensified precipitationstrengthens the Hadley and Walker circulations inthe troposphere, with an associated increase intrade wind strength that produces greater equa-torial ocean upwelling and lower equatorial SSTsin the eastern Pacific, a signal that was firstdiscovered in observational data (1, 2). The en-hanced subsidence produces fewer clouds in theequatorial eastern Pacific and the expanded sub-tropical regions that allow even more solar ra-diation to reach the surface to produce a positivefeedback (18, 19). Dynamical air-sea couplingproduces a transition to higher eastern equatorialSSTs a couple of years later (20, 21). There isobservational evidence for a strengthenedHadleycirculation in peak solar forcing years associatedwith intensified tropical precipitation maxima, astronger descending branch in the subtropics, anda stronger ascending branch in the lower latitudes(3); a poleward expansion of the Hadley circula-tion in peak solar years, with stronger ascendingmotions at the edge of the rising branch, as wellas a stronger Walker circulation with enhancedupward motions in the tropical western Pacificconnected to stronger descending motions in thetropical eastern Pacific (7); and enhanced sum-mer season off-equatorial climatological monsoonprecipitation over India (6, 22). This cold event–like response to peak solar forcing is differentfrom cold events (also known as La Niña events)

1National Center for Atmospheric Research, Post Office Box3000 Boulder, CO 80307, USA. 2Center for AustralianWeather and Climate, Bureau of Meteorology, Melbourne,Australia. 3Helmholtz Centre Potsdam – GFZ German ResearchCentre for Geosciences, Potsdam, Germany. 4Institut fürMeteorologie, Freie Universität Berlin, Berlin, Germany.5Naval Research Laboratory, Washington, DC 20375, USA.6Colorado Research Associates, Boulder, CO 80301, USA.

*To whom correspondence should be addressed. E-mail:meehl@ncar.ucar.edu

28 AUGUST 2009 VOL 325 SCIENCE www.sciencemag.org1114

REPORTS

on

Sep

tem

ber

13, 2

010

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om